Browse Comments

Sorry, AI is too busy boiling the coral reef to call Elon Musk a pedophile.

No need to worry... this entire research field is basically full of shit. Or, to put it another way: there is no fucking chance in hell that all this research will result in anything capable of "aligning" even basic intelligence. how should aligning human level intelligence work then? But I'll let this thread express what I want to say, with much more dignity and less f-words: https://www.lesswr…

A lot of people seem to be pointing out flaws with deep research but this is to be expected. Forget about company specific AI research products. Zoom out and look at the big picture. This is the worst it’s ever going to be. It will eventually severely outpace human knowledge in every domain.

Open source finally catching up to the voice to voice models. I've mostly only been seeing TTS and STT. That being said the internal model for Sesame is quite small, so it's nowhere near "intelligent" as GPT 4o, nor is it actually fully multi modal. Now consider the fact that OpenAI had GPT 4o internally around a year ago. The completely uncensored version. We know how good it is based on their …

I mean, if you want I could just make up a list of issues if that’s what you’re after. There’s no penalty for AI being wrong. Doctors have an obligation to actually diagnose you correctly. Those 5 years of specialists and asking chat GPT some questions are not even remotely comparable and you should absolutely not trust what AI says in terms of medical advice.

Yup. In this metaphor, which I love love love, I'd argue Nvidia is the only one selling shovels. Zuck and SA are still trying to sell gold/fools gold they mined up. Zuck, less clear precisely what he's selling, but SA is trying to sell gold. Another shovel seller: ASML. Also, AMAT selling pick-axes as a shovel augmenter in this gold rush: https://www.youtube.com/watch?v=GSuTyOMq1Bg I love …

This is the canary in the coal mine. STEM professions were the last attainable, well paying white-collar professions. The tech sector is around 10% of the GDP itself and it is obviously hurting, there's huge credential inflation for Software Developers, rapid outsourcing, and opportunities are dwindling. Medical professions are ok cause there's a lot of aging boomers... But the ladder is being pu…

Remember that movie Her, about a guy who developed an unhealthy relationship with an AI? Apparently it's happening in real life now, and Sam Altman counts it among the "cool" uses for ChatGPT. At the Sequoia Capital AI Ascent conference earlier this month, OpenAI CEO Sam Altman answered a slew of questions about ChatGPT, including a few that homed in on some of the seriously concerning trends ar…

I replied elsewhere in the comments, but I think it should be said again: The ‘neural howlround’ paper was written by a solo author with no affiliation or credentials and was published without peer review. The paper has only 2 references, it is without any underlying basis. Attempting to google for the author, Seth Drake, produces no useful results. Drake thanks Professor Jim Hendler from RPI i…

Yeah, it’s a distinct possibility, and I think a lot of reddit is delusional in both directions. I’m not sure AI will kill all humans in 5 years. I’m also not sure AI is just a new helpful utility. I’m concerned that a big segment of our economy and approach to higher ed is built around jobs that could legitimately be done by AI at a pretty comparable level for a fraction of the cost, and those j…

They won't admit AI isn't as good as the hype until it blows up in their faces

People here are so obsessed with ChatGPT that when these stories of delusion come up, they always claim the posters are lying. But if you take ChatGPT in even a moderately spiritual/philosophical direction, it will absolutely get esoteric on you. It’s not hard at all to get it going.

Yeah, my optimism stems from AI remaining stuck in the limitations of it being a language model, not our humanity. It feels more solid.

It's not that nobody *knows* what they are, it's that nobody *agrees* what they are. Language works because words have agreed-upon meanings. When there isn't an agreed-upon meaning language fails in its role to communicate ideas between people. Person 1 can say "Chat GPT isn't even AI," Person 2 can say "ChatGPT is obviously AI, but it obviously isn't AGI," Person 3 can say "ChatGPT is AGI, but …

Aw geez, I feel like there’s already a ton of AI-generated images hidden in my search results.

There's no pattern other than the fact that it isn't possible to detect AI generated images unless they label themselves as such.

It’s a glorified language model. Come back to me when it actually does something other than make people dumber or [drive them to madness](https://futurism.com/openai-investor-chatgpt-mental-health).

Noob end user here in regards to chatgpt but I always assumed the older models were more outdated and were too specific of use-case. I never dabbled with them as I figured those were just archives of older generations. Personally, for what I use it for, I'm happy to see 5 naturally reduce the twitter emoji feel and its more concise. Again, however, I'm not an AI power user.

This. Frankly we are doing testing on every Enterprise AI tool against our codebases and we are finding that it will do trivial tasks but anything with a large existing codebase, in our case python / java, it causes way more tech debt and issues then if someone half assed a ticket. Don’t get me wrong, copilot PR review is nifty but pretty inaccurate. The power in AI is around researching quick…

>How do we get from LLMs to AGI? We don't. At least not directly. As you said, LLMs and AGI are vastly different things. There's no clear path from LLM to AGi. It's not a matter of processing power or algorithm optimization. It will require completely new technologies we haven't even created yet. It's like asking how we go from a bicycle to an interstellar spaceship. I'm not naive enou…